CN112214962B - Circuit original input sensitivity calculation method based on probability model - Google Patents

Abstract

A circuit original input sensitivity calculation method based on probability model, netlist analysis and related quantity initialization; judging the criticality Crtl _i, j of the j-th input end of the i-th node in the LC; pruning the circuit according to the result to obtain circuit sensitivity nodes corresponding to the applied input vector, and assigning sensitivity values to the marked sensitivity nodes; and (3) giving out the number of clusters of sen based on the kernel density estimation, clustering sen by using a k-means algorithm, and outputting a clustering result according to the order of the sensitivity values. The pruning strategy based on the shielding effect realizes the marking of the sensitive original input end under the given input vector; the quantification of the sensitivity level of each original input end of the circuit is realized by means of SCA algorithm based on the reverse depth-first search strategy through the type and the topological position information of the nodes; and by combining a clustering algorithm, effective identification of the sensitive original input end in the circuit is realized through continuous iteration.

Description

Circuit original input sensitivity calculation method based on probability model

Technical Field

The invention relates to sensitivity calculation of an original input end of an integrated circuit, in particular to a calculation method based on combination of pruning strategy and shielding effect measurement.

Background

The ever shrinking feature sizes of integrated circuits inevitably results in reduced circuit reliability margins. Studies have shown that the reliability of a circuit is affected by the input vector of the circuit, and that the reliability differences at different input vectors sometimes differ even by several orders of magnitude. For this reason, in the circuit design process, it is necessary to know the boundaries of circuit reliability and their corresponding input vectors so as to purposefully improve the reliability level of the circuit. Developing an analysis of the sensitivity of the original input of the circuit is an effective way to achieve the above objectives.

Currently, a learner based on intelligent algorithms to evaluate the sensitivity of the original inputs of the circuit. Typically, a hill climbing algorithm is used in combination with a recursive search to find the worst input vector for the circuit; there are also methods for determining the sensitivity of the original input of the circuit by machine learning based on TetraMAX collected data. However, the above method has a large computational overhead when facing a circuit with a complex structure, and the self-learning ability of the algorithm is still to be improved to further improve the computational accuracy.

Disclosure of Invention

In order to overcome the defects of the prior art and realize effective identification of the sensitive original input end of a large-scale circuit, the invention provides a calculation method of the sensitive original input end of the circuit based on a probability model, and a pruning strategy based on a shielding effect. And then, based on a reverse depth-first search strategy and by means of SCA algorithm, the sensitivity level of each original input end of the circuit is quantized through the type and topological position information of the nodes. The clustering algorithm is combined, and the effective identification of the sensitive original input end in the circuit can be realized through continuous iteration of the steps.

The technical scheme adopted for solving the technical problems is as follows:

A method for calculating sensitivity of an original input end of a circuit based on a probability model, the method comprising the steps of:

step 1: the process of netlist parsing and related quantity initialization is as follows:

1.1 The method comprises the steps of) reading a netlist, extracting basic gate information of a circuit, constructing an integrity linked list LC of the corresponding circuit, and identifying all original input ends and original output ends of the circuit, wherein the linked list LC refers to any node of which the input end information can be extracted from the output end information of a preamble node of the node;

1.2 According to the type of the node, the fault probability p and the number m of the input ends, constructing a probability transition matrix PM _type-m and an ideal transition matrix IM _type-m of the proposed node based on a true value table method; setting a cyclic variable k=1;

1.3 Judging whether k is smaller than the designated calculation times, if not, transposing the step 4; extracting input signals of all original input ends of the circuit, and constructing original input probability distribution pipt corresponding to the input signals; when the original input signal is 0, the corresponding pipt is assigned a value of [1,0]; otherwise pipt is assigned a value of [0,1]; initializing i=1;

Step 2: the process of determining the criticality Crtl _i,j of the j-th input of the i-th node in LC is as follows:

2.1 If i is more than or equal to Length (LC), turning to step 3; otherwise, go to 2.2); wherein length is used to calculate the length of the sequence;

2.2 Acquiring a current original input signal of the circuit, and calling a SCA algorithm to acquire an ideal output probability signal iopt _i of an ith node g _i and an ideal input probability signal iipt _i,j of a jth input end of the ideal output probability signal; where j=1, 2, …, mi, mi refers to the number of inputs to node i; reset j=0;

2.3 If j= mi, then perform i=i+1 and go to step 2.1); otherwise, perform j=j+1 and go 2.4);

2.4 Setting iipt _i,j again to calculate the output probability signal tfopt _i,j of gi in the current case;

2.5 If tfopt _i,j＝＝iopt_i, crtl _i,j =0 is performed; otherwise, crtl _i,j =1 is performed; wherein Crtl _i,j denotes the sensitivity state of the j-th input of the i-th node;

Step 3: pruning the circuit according to the result of the step 2 to obtain circuit sensitivity nodes corresponding to the applied input vectors, and assigning sensitivity values to the marked sensitivity nodes;

step 4: and (3) giving out the number of clusters of sen based on the kernel density estimation, clustering sen by using a k-means algorithm, and outputting a clustering result according to the order of the sensitivity values.

Further, the procedure of the step 3 is as follows:

3.1 A) invoking a SCA algorithm to obtain a current output probability signal fopt _i, i=1, 2, …, length (LC) for gi at the current applied input vector; initializing loop variables t=1 and i=length (LC);

3.2 Reading gi), extracting a probability transition matrix PM _type-mi corresponding to the gi, an ideal transition matrix IM _type-mi, an output probability signal fopt _i, an ideal output signal iopt _i and the number mi of input ends of the ideal output signal iopt _i, executing visited _i =1, and calculating a sensitive value CrtVl _i＝sum(fopt_i.×iopt_i), wherein visited _i refers to the accessed state of the node gi, and sum is used for obtaining the sum of elements in the vector;

3.3 Extracting the type of gi and the number num of key input ends thereof;

3.4 If num= =0, let PERTEMCRTVL _i,t =0 and turn 3.13), where t=1, 2, …, mi, PERTEMCRTVL _i,t represents the sensitivity percentage of the t-th input of the i-th node;

3.5 If num= =1, let PERTEMCRTVL _i,t =1 and turn 3.13), where t is the corresponding unique critical input sequence number;

3.6 If num >1 and type is xor, or buff, or not, then PERTEMCRTVL _i,t =1/num is executed and 3.13) is turned, where t represents the critical input number corresponding to the node;

3.7 If num >1 and type is not xor, buff, and not, sequentially placing all the key input end sequence numbers of the node into the vector keyipt;

3.8 Extracting tensor product operation results GIp of output probability signals corresponding to all elements in keyipt, and initializing a cyclic variable p=1;

3.9 If p < = length (GIp), then p-1 is expressed by a power of 2 and a form, and the power and the coefficient are put into a vector iptnow as a power corresponding to 1; otherwise, turning to 3.13);

3.10 Initializing a loop variable q=1; if the type is and or nand, turning to 3.11); otherwise, turning to 3.12);

3.11 If q < = length (keyipt), judging whether ismember (q-1, iptnow) = 1 is true, if not, executing q=q+1 and converting 3.11), if true, calculating PERTEMCRTVL _i,q by using the formula (2) and the formula (3), and executing q=q+1 and converting 3.11); otherwise, q=q+1 and 3.9 is performed; wherein ismember is used for judging whether the left parameter is an element in the right parameter, PERTEMCRTVL is a temporary variable;

pertemCrtVl＝sum((GIp(p).×PM_type-m(p,:)/(length(iptnow))).×IM_type-mi)/CrtVl_p (2)

pertemCrtVl_i,q＝pertemCrtVl_i,q+pertemCrtVl (3)

3.12 If q < = length (keyipt), judging whether ismember (q-1, iptnow) = 0 is true, if not, executing q=q+1 and converting 3.12), if true, calculating PERTEMCRTVL _i,q by using the formula (4) and the formula (5), and executing q=q+1 and converting 3.12); otherwise, q=q+1 and 3.9 is performed;

pertemCrtVl＝sum((GIp(p).×PM_type-m(p,:)/length(keyipt)-length(iptnow)).×IM_type-mi)/CrtVl_p (4)

pertemCrtVl_p,j＝pertemCrtVl_p,j+pertemCrtVl (5)

3.13 If the type of gi is not the original input, calculate TRANSCRTVL _i,q as per equation (6), otherwise calculate TRANSCRTVL _i,q as per equation (7); wherein TRANSCRTVL _i denotes the sensitivity value corresponding to the gi output terminal, TRANSCRTVL _i,q denotes the sensitivity value of the q-th critical input terminal of gi;

transCrtVl_i,q＝transCrtVl_i*pertemCrtVl_i,q (6)

transCrtVl_i,q＝transCrtVl_i,q+transCrtVl_i*pertemCrtVl_i,q (7)

3.14 Initializing q=1;

3.15 If q > mi), turn 3.16); otherwise, extracting the q-th input port ipt _i,q of gi, and turning to 3.16);

3.16 If ipt _i,q is the key input terminal and visited _i,q =0, updating i by using the number corresponding to the input terminal q, and converting to 3.2); otherwise, q=q+1 and 3.15 is performed; wherein visited _i,q denotes the accessed state of the q-th input of gi;

3.17 Respectively counting the accumulated sensitivity values of the original input ends, placing the accumulated sensitivity values in a set sen, and then turning to 1.3).

The technical conception of the invention is as follows: first, for the applied input vector, the sensitive inputs of each node are identified based on the masking effect of the circuit. Then, through the sensitive input end of each node, the node type and the position information in the circuit topology, the sensitivity level of the relevant input end is quantized to the original input end of the circuit by utilizing an inverse depth-first search strategy and combining with the SCA algorithm. Then, new input vectors are continuously randomly generated, and the steps are repeatedly executed until the end condition is reached. And finally, summarizing the sensitivity values of the original input ends under different input vectors, and realizing the measurement of the sensitivity level of the original input ends of the circuit based on a clustering method. The result is helpful for circuit designers to know and master the sensitivity level of each original input end of the circuit in the designed product in time so as to reasonably select and decide.

The beneficial effects of the invention are mainly shown in the following steps: the method is beneficial to quickly and effectively identifying the sensitive input end of the circuit by mistake, so that the test of the reliability of the circuit structure is quickened in the early stage of circuit design, and the upper and lower bounds of the reliability of the circuit structure, the possible application environment thereof and the like are timely defined.

Drawings

FIG. 1 is a flow chart of a method for computing sensitivity of an original input of a circuit based on a probabilistic model.

Detailed Description

The invention is further described below with reference to the accompanying drawings.

Referring to fig. 1, a method for calculating sensitivity of an original input end of a circuit based on a probability model includes the following steps:

step 1: the process of netlist parsing and related quantity initialization is as follows:

1.1 The method comprises the steps of) reading a netlist, extracting basic gate information of a circuit, constructing an integrity linked list LC of the corresponding circuit, and identifying all original input ends and original output ends of the circuit, wherein the linked list LC refers to any node of which the input end information can be extracted from the output end information of a preamble node of the node;

1.2 According to the type of the node, the fault probability p and the number m of the input ends, constructing a probability transition matrix PM _type-m and an ideal transition matrix IM _type-m of the proposed node based on a true value table method; setting a cyclic variable k=1;

1.3 Judging whether k is smaller than the designated calculation times, if not, transposing the step 4; extracting input signals of all original input ends of the circuit, and constructing original input probability distribution pipt corresponding to the input signals; when the original input signal is 0, the corresponding pipt is assigned a value of [1,0]; otherwise pipt is assigned a value of [0,1]; initializing i=1;

Step 2: the process of determining the criticality Crtl _i,j of the j-th input of the i-th node in LC is as follows:

2.1 If i is more than or equal to Length (LC), turning to step 3; otherwise, go to 2.2); wherein length is used to calculate the length of the sequence;

2.2 Acquiring a current original input signal of the circuit, and calling a SCA algorithm to acquire an ideal output probability signal iopt _i of an ith node g _i and an ideal input probability signal iipt _i,j of a jth input end of the ideal output probability signal; where j=1, 2, …, mi, mi refers to the number of inputs to node i; reset j=0;

2.3 If j= mi, then perform i=i+1 and go to step 2.1); otherwise, perform j=j+1 and go 2.4);

2.4 Setting iipt _i,j again to calculate the output probability signal tfopt _i,j of gi in the current case;

2.5 If tfopt _i,j＝＝iopt_i, crtl _i,j =0 is performed; otherwise, crtl _i,j =1 is performed; wherein Crtl _i,j denotes the sensitivity state of the j-th input of the i-th node.

Step 3: pruning the circuit according to the result of step 2 to obtain circuit sensitivity nodes corresponding to the applied input vectors, and assigning sensitivity values to the labeled sensitivity nodes.

3.1 A) invoking a SCA algorithm to obtain a current output probability signal fopt _i, i=1, 2, …, length (LC) for gi at the current applied input vector; initializing loop variables t=1 and i=length (LC);

3.2 Reading gi), extracting a probability transition matrix PM _type-mi corresponding to the gi, an ideal transition matrix IM _type-mi, an output probability signal fopt _i, an ideal output signal iopt _i and the number mi of input ends of the ideal output signal iopt _i, executing visited _i =1, and calculating a sensitive value CrtVl _i＝sum(fopt_i.×iopt_i), wherein visited _i refers to the accessed state of the node gi, and sum is used for obtaining the sum of elements in the vector;

3.3 Extracting the type of gi and the number num of key input ends thereof;

3.4 If num= =0, let PERTEMCRTVL _i,t =0 and turn 3.13), where t=1, 2, …, mi, PERTEMCRTVL _i,t represents the sensitivity percentage of the t-th input of the i-th node;

3.5 If num= =1, let PERTEMCRTVL _i,t =1 and turn 3.13), where t is the corresponding unique critical input sequence number;

3.6 If num >1 and type is xor, or buff, or not, then PERTEMCRTVL _i,t =1/num is executed and 3.13) is turned, where t represents the critical input number corresponding to the node;

3.7 If num >1 and type is not xor, buff, and not, sequentially placing all the key input end sequence numbers of the node into the vector keyipt;

3.8 Extracting tensor product operation results GIp of output probability signals corresponding to all elements in keyipt, and initializing a cyclic variable p=1;

3.9 If p < = length (GIp), then p-1 is expressed by a power of 2 and a form, and the power and the coefficient are put into a vector iptnow as a power corresponding to 1; otherwise, turning to 3.13);

3.10 Initializing a loop variable q=1; if the type is and or nand, turning to 3.11); otherwise, turning to 3.12);

3.11 If q < = length (keyipt), judging whether ismember (q-1, iptnow) = 1 is true, if not, executing q=q+1 and converting 3.11), if true, calculating PERTEMCRTVL _i,q by using the formula (2) and the formula (3), and executing q=q+1 and converting 3.11); otherwise, q=q+1 and 3.9 is performed; wherein ismember is used for judging whether the left parameter is an element in the right parameter, PERTEMCRTVL is a temporary variable;

pertemCrtVl＝sum((GIp(p).×PM_type-m(p,:)/(length(iptnow))).×IM_type-mi)/CrtVl_p (2)

pertemCrtVl_i,q＝pertemCrtVl_i,q+pertemCrtVl (3)

3.12 If q < = length (keyipt), judging whether ismember (q-1, iptnow) = 0 is true, if not, executing q=q+1 and converting 3.12), if true, calculating PERTEMCRTVL _i,q by using the formula (4) and the formula (5), and executing q=q+1 and converting 3.12); otherwise, q=q+1 and 3.9 is performed;

pertemCrtVl＝sum((GIp(p).×PM_type-m(p,:)/length(keyipt)-length(iptnow)).×IM_type-mi)/CrtVl_p (4)

pertemCrtVl_p,j＝pertemCrtVl_p,j+pertemCrtVl (5)

3.13 If the type of gi is not the original input, calculate TRANSCRTVL _i,q as per equation (6), otherwise calculate TRANSCRTVL _i,q as per equation (7); wherein TRANSCRTVL _i denotes the sensitivity value corresponding to the gi output terminal, TRANSCRTVL _i,q denotes the sensitivity value of the q-th critical input terminal of gi;

transCrtVl_i,q＝transCrtVl_i*pertemCrtVl_i,q (6)

transCrtVl_i,q＝transCrtVl_i,q+transCrtVl_i*pertemCrtVl_i,q (7)

3.14 Initializing q=1;

3.15 If q > mi), turn 3.16); otherwise, extracting the q-th input port ipt _i,q of gi, and turning to 3.16);

3.16 If ipt _i,q is the key input terminal and visited _i,q =0, updating i by using the number corresponding to the input terminal q, and converting to 3.2); otherwise, q=q+1 and 3.15 is performed; wherein visited _i,q denotes the accessed state of the q-th input of gi;

3.17 Respectively counting the accumulated sensitivity values of the original input ends, placing the accumulated sensitivity values in a set sen, and then turning to 1.3);

step 4: and (3) giving out the number of clusters of sen based on the kernel density estimation, clustering sen by using a k-means algorithm, and outputting a clustering result according to the order of the sensitivity values.

The embodiment is based on a pruning strategy facing shielding effect, combines an SCA algorithm, and utilizes a probability method to realize effective quantification of the sensitivity level of the original input end of the circuit through a reverse recursion method and the type and position information of each node. This will play an important role in the fast and efficient quantization of the reliability boundaries of the circuit structure.

The embodiments described in the present specification are merely examples of implementation forms of the inventive concept, and the scope of protection of the present invention should not be construed as being limited to the specific forms set forth in the embodiments, and the scope of protection of the present invention and equivalent technical means that can be conceived by those skilled in the art based on the inventive concept.

Claims (2)

1. The method for calculating the sensitivity of the original input end of the circuit based on the probability model is characterized by comprising the following steps:

step 1: the process of netlist parsing and related quantity initialization is as follows:

1.1 The method comprises the steps of) reading a netlist, extracting basic gate information of a circuit, constructing an integrity linked list LC of the corresponding circuit, and identifying all original input ends and original output ends of the circuit, wherein the linked list LC refers to any node of which the input end information can be extracted from the output end information of a preamble node of the node;

1.2 According to the type of the node, the fault probability p and the number m of the input ends, constructing a probability transition matrix PM _type-m and an ideal transition matrix IM _type-m of the proposed node based on a true value table method; setting a cyclic variable k=1;

1.3 Judging whether k is smaller than the designated calculation times, if not, transposing the step 4; extracting input signals of all original input ends of the circuit, and constructing original input probability distribution pipt corresponding to the input signals; when the original input signal is 0, the corresponding pipt is assigned a value of [1,0]; otherwise pipt is assigned a value of [0,1]; initializing i=1;

Step 2: the process of determining the criticality Crtl _i,j of the j-th input of the i-th node in LC is as follows:

2.1 If i is more than or equal to Length (LC), turning to step 3; otherwise, go to 2.2); wherein length is used to calculate the length of the sequence;

2.2 Acquiring a current original input signal of the circuit, and calling a SCA algorithm to acquire an ideal output probability signal iopt _i of an ith node g _i and an ideal input probability signal iipt _i,j of a jth input end of the ideal output probability signal; where j=1, 2, …, mi, mi refers to the number of inputs to node i; reset j=0;

2.3 If j= mi, then perform i=i+1 and go to step 2.1); otherwise, perform j=j+1 and go 2.4);

2.4 Setting iipt _i,j again to calculate the output probability signal tfopt _i,j of gi in the current case;

2.5 If tfopt _i,j＝＝iopt_i, crtl _i,j =0 is performed; otherwise, crtl _i,j =1 is performed; wherein Crtl _i,j denotes the sensitivity state of the j-th input of the i-th node;

Step 3: pruning the circuit according to the result of the step 2 to obtain circuit sensitivity nodes corresponding to the applied input vectors, and assigning sensitivity values to the marked sensitivity nodes;

step 4: and (3) giving out the number of clusters of sen based on the kernel density estimation, clustering sen by using a k-means algorithm, and outputting a clustering result according to the order of the sensitivity values.

2. The method for calculating the sensitivity of the original input end of the circuit based on the probability model as set forth in claim 1, wherein the process of the step 3 is as follows:

3.1 A) invoking a SCA algorithm to obtain a current output probability signal fopt _i, i=1, 2, …, length (LC) for gi at the current applied input vector; initializing loop variables t=1 and i=length (LC);

3.2 Reading gi), extracting a probability transition matrix PM _type-mi corresponding to the gi, an ideal transition matrix IM _type-mi, an output probability signal fopt _i, an ideal output signal iopt _i and the number mi of input ends of the ideal output signal iopt _i, executing visited _i =1, and calculating a sensitive value CrtVl _i＝sum(fopt_i.×iopt_i), wherein visited _i refers to the accessed state of the node gi, and sum is used for obtaining the sum of elements in the vector;

3.3 Extracting the type of gi and the number num of key input ends thereof;

3.4 If num= =0, let PERTEMCRTVL _i,t =0 and turn 3.13), where t=1, 2, …, mi, PERTEMCRTVL _i,t represents the sensitivity percentage of the t-th input of the i-th node;

3.5 If num= =1, let PERTEMCRTVL _i,t =1 and turn 3.13), where t is the corresponding unique critical input sequence number;

3.6 If num >1 and type is xor, or buff, or not, then PERTEMCRTVL _i,t =1/num is executed and 3.13) is turned, where t represents the critical input number corresponding to the node;

3.7 If num >1 and type is not xor, buff, and not, sequentially placing all the key input end sequence numbers of the node into the vector keyipt;

3.8 Extracting tensor product operation results GIp of output probability signals corresponding to all elements in keyipt, and initializing a cyclic variable p=1;

3.9 If p < = length (GIp), then p-1 is expressed by a power of 2 and a form, and the power and the coefficient are put into a vector iptnow as a power corresponding to 1; otherwise, turning to 3.13);

3.10 Initializing a loop variable q=1; if the type is and or nand, turning to 3.11); otherwise, turning to 3.12);

3.11 If q < = length (keyipt), judging whether ismember (q-1, iptnow) = 1 is true, if not, executing q=q+1 and converting 3.11), if true, calculating PERTEMCRTVL _i,q by using the formula (2) and the formula (3), and executing q=q+1 and converting 3.11); otherwise, q=q+1 and 3.9 is performed; wherein ismember is used for judging whether the left parameter is an element in the right parameter, PERTEMCRTVL is a temporary variable;

pertemCrtVl＝sum((GIp(p).×PM_type-m(p,:)/(length(iptnow))).×IM_type-mi)/CrtVl_p (2)

pertemCrtVl_i,q＝pertemCrtVl_i,q+pertemCrtVl (3)

3.12 If q < = length (keyipt), judging whether ismember (q-1, iptnow) = 0 is true, if not, executing q=q+1 and converting 3.12), if true, calculating PERTEMCRTVL _i,q by using the formula (4) and the formula (5), and executing q=q+1 and converting 3.12); otherwise, q=q+1 and 3.9 is performed;

pertemCrtVl＝sum((GIp(p).×PM_type-m(p,:)/length(keyipt)-length(iptnow)).×IM_type-mi)/CrtVl_p (4)

pertemCrtVl_p,j＝pertemCrtVl_p,j+pertemCrtVl (5)

3.13 If the type of gi is not the original input, calculate TRANSCRTVL _i,q as per equation (6), otherwise calculate TRANSCRTVL _i,q as per equation (7); wherein TRANSCRTVL _i denotes the sensitivity value corresponding to the gi output terminal, TRANSCRTVL _i,q denotes the sensitivity value of the q-th critical input terminal of gi;

transCrtVl_i,q＝transCrtVl_i*pertemCrtVl_i,q (6)

transCrtVl_i,q＝transCrtVl_i,q+transCrtVl_i*pertemCrtVl_i,q (7)

3.14 Initializing q=1;

3.15 If q > mi), turn 3.16); otherwise, extracting the q-th input port ipt _i,q of gi, and turning to 3.16);

3.16 If ipt _i,q is the key input terminal and visited _i,q =0, updating i by using the number corresponding to the input terminal q, and converting to 3.2); otherwise, q=q+1 and 3.15 is performed; wherein visited _i,q denotes the accessed state of the q-th input of gi;

3.17 Respectively counting the accumulated sensitivity values of the original input ends, placing the accumulated sensitivity values in a set sen, and then turning to 1.3).